Jumping retroviruses nudge TADs apart

Far from being junk DNA, the pervasive retrotransposons that populate the genome have a powerful capacity to influence genes and chromatin. A new study demonstrates how the transcription of one such element, HERV-H, can modify the higher-order 3D structure of chromatin during early primate development.

Constituting roughly half of human DNA, retrotransposons such as endogenous retroviruses (ERVs) have the extraordinary ability to jump and propagate throughout the genome (Fig. 1). Despite being silenced during development by repressive epigenetic marks, during evolution, many ERVs have been co-opted for critical cellular functions. For example, ERVs frequently act to distribute regulatory information and thus confer genes with new patterns of expression and function1. Similarly, multiple ERVs have been re-purposed as novel genes themselves, including the ERV envelope-derived syntactins that drive trophoblast fusion and establish the placental fetal–maternal interface2. Now, Zhang et al. have identified an additional activity of co-opted ERVs—the alteration of the genome’s 3D structure itself—that may have the capacity to substantially affect gene regulation3.

Fig. 1: Functionally co-opted ERVs have diverse activities.

a, Expression from Pol II promoters in ERV long terminal repeats (LTRs) allows their ‘copy and paste’ transposition throughout the genome until epigenetic silencing. KRAB-ZFP, KRAB zinc-finger protein. b, Newly inserted ERVs are frequently co-opted, for example, to disrupt or create genes as well as to alter their regulation by distributing transcription factor (TF)- or insulator (CTCF)-binding sites. c, Transcription of HERV-Hs creates an hESC-specific boundary at the ERVs 3′ end that breaks a TAD in two, perhaps by blocking cohesin from extruding chromatin in one direction.

Splitting TADs with transcription

The past decade has revealed that chromatin folding spatially partitions the genome in 3D space into many discrete self-interacting blocks termed topologically associating domains (TADs)4. It is now clear that partitioning into TADs refines transcriptional regulation by physically isolating promoters together with cis-regulatory enhancer elements that drive their correct expression pattern5. In mammals, TADs appear to be formed by a process of loop extrusion, whereby chromatin loops are progressively pulled by cohesin through its ring structure until stalling occurs at extrusion-blocking boundaries6. Most TAD boundaries are critical to setting the limits of extrusion; they are invariant and thus maintain TADs across different species and cell types. However, some boundaries are cell-type specific and consequently may dynamically influence gene regulation7. Nevertheless, what distinguishes these variable boundaries from their permanent counterparts remains largely unknown.

To identify the features defining cell-type-specific TAD boundaries, Zhang et al. have examined changes in chromatin structure occurring during in vitro human cardiac differentiation. Similarly to findings from previous reports, the authors observed cell-type-specific boundaries at each stage7. However, the boundaries unique to human embryonic stem cells (hESCs) were enriched in eight classes of repeat elements, including the primate-specific HERV-H. Unexpectedly, the capacity for the HERV-H elements to act as boundaries appeared to be linked to their transcription; of ~1,000 human HERV-H insertions, only the top 50 most expressed had boundary activities. Zhang et al. compellingly tested this transcriptional dependency and found that boundary activities were lost when HERV-H elements were silenced naturally during differentiation and artificially through CRISPR interference in hESCs (Fig. 1). Moreover, this capacity for boundary generation was maintained when HERV was transcribed in pluripotent cells of other primates. Thus, boundaries between TADs can be created by transcription of primate-specific ERVs.

The authors then investigated how ERV transcription drives insulation by examining the distribution of the TAD-forming extruding factor cohesin and CTCF, a protein that defines most TAD boundaries8. Despite not displaying differential CTCF binding that might account for their hESC specificity, active HERV-H boundaries did show an unexpected accumulation of cohesin at the ERV 3′ ends. Interestingly, RNA polymerase II (Pol II) is known to push cohesin in the direction of transcriptional elongation and consequently can drive cohesin accumulation at the 3′ ends of genes9. Thus, the boundary activity of active HERV-H may derive from robust transcription mono-directionally blocking extrusion complexes moving against the direction of elongation (Fig. 1c).

If this were the case, would all highly transcribed genes possess boundary activity? Indeed, 6% of TAD boundaries appear to be defined solely by highly transcribed housekeeping genes4. However, after examining the correlation between insulation and transcription, Zhang et al. did not observe significant boundary activity in the top 1,000 most expressed genes. Thus, transcription-dependent boundaries are likely to depend on yet-unknown features unique to the promoters of active ERVs and perhaps boundary-associated housekeeping genes. Interestingly, transcription is now known to occur in stochastic bursts of elongating Pol II (ref. 10). Thus, one intriguing possibility is that promoters at transcription-dependent boundaries may display higher frequencies of bursting that fire the constant stream of Pol II necessary to consistently block cohesin extrusion.

Regulatory and disease consequences

These ERV-driven structural changes have unknown consequences in gene regulation and disease. In the study by Zhang et al., deletion or repression of two HERV-H boundaries decreased the expression of upstream genes within the same TAD, thus suggesting that the elements function as enhancers as well as barriers. Nevertheless, similar ERV-transcription-dependent TAD boundaries with positions varying among mouse strains have recently been reported in two-cell-stage mouse embryos11. Hence, random ERV transposition can seemingly alter chromatin structure differently across distinct individuals. Because the introduction of boundaries can separate promoters from their enhancers, such varied transposition events may have considerable but individual-specific disruptive and pathogenic effects on gene expression12. However, the ERV-dependent boundaries observed to date appear to be active only in pluripotent cells and to disappear during differentiation. Thus, it is unclear whether such ERV-dependent boundaries might also be active in the non-pluripotent cell types necessary to drive pathogenic gene mis-expression in later development. Unfortunately, clarifying this possibility will not be easy. The repetitive nature of ERVs makes their de novo identification by sequencing challenging across individuals, although new long-read technologies may greatly help13. Likewise, although boundary formation is dependent on transcription, the factors driving boundary formation and thus the cell types in which it occurs are largely unknown. Consequently, although ERVs possess a potent mechanism to influence chromatin structure, determining the relevance of this mechanism to normal and pathogenic gene regulation will be a major but exciting challenge.


  1. 1.

    Chuong, E. B., Elde, N. C. & Feschotte, C. Nat. Rev. Genet. 18, 71–86 (2017).

    CAS  Article  Google Scholar 

  2. 2.

    Mi, S. et al. Nature 403, 785–789 (2000).

    CAS  Article  Google Scholar 

  3. 3.

    Zhang, Y. et al. Nat. Genet. https://doi.org/10.1038/s41588-019-0479-7 (2019).

    CAS  Article  Google Scholar 

  4. 4.

    Dixon, J. R. et al. Nature 485, 376–380 (2012).

    CAS  Article  Google Scholar 

  5. 5.

    Robson, M. I., Ringel, A. R. & Mundlos, S. Mol. Cell 74, 1110–1122 (2019).

    CAS  Article  Google Scholar 

  6. 6.

    Fudenberg, G. et al. Cell Rep. 15, 2038–2049 (2016).

    CAS  Article  Google Scholar 

  7. 7.

    Bonev, B. et al. Cell 171, 557–572.e24 (2017).

    CAS  Article  Google Scholar 

  8. 8.

    Nora, E. P. et al. Cell 169, 930–944.e22 (2017).

    CAS  Article  Google Scholar 

  9. 9.

    Busslinger, G. A. et al. Nature 544, 503–507 (2017).

    CAS  Article  Google Scholar 

  10. 10.

    Nicolas, D., Phillips, N. E. & Naef, F. Mol. Biosyst. 13, 1280–1290 (2017).

    CAS  Article  Google Scholar 

  11. 11.

    Kruse, K. et al. Preprint at bioRxiv https://doi.org/10.1101/523712 (2019).

  12. 12.

    Despang, A. et al. Nat. Genet. https://doi.org/10.1038/s41588-019-0466-z (2019).

    CAS  Article  Google Scholar 

  13. 13.

    Mantere, T., Kersten, S. & Hoischen, A. Front. Genet. 10, 426 (2019).

    CAS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Stefan Mundlos.

Ethics declarations

Competing interests

The authors declare no competing interests.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Robson, M.I., Mundlos, S. Jumping retroviruses nudge TADs apart. Nat Genet 51, 1304–1305 (2019). https://doi.org/10.1038/s41588-019-0491-y

Download citation


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing